seqlearn

https://github.com/larsmans/seqlearn

"""
Generic sequence prediction script using CoNLL format.
"""

from __future__ import print_function
import fileinput
from glob import glob
import sys

from seqlearn.datasets import load_conll
from seqlearn.evaluation import bio_f_score
from seqlearn.perceptron import StructuredPerceptron
from sklearn.metrics import accuracy_score
from sklearn.externals import joblib


if __name__ == "__main__":
    X_train, y_train, lengths_train = load_conll(sys.argv[1], features, 100000, True)
    #describe(X_train, lengths_train)

    X_test, y_test, lengths_test = load_conll(sys.argv[2], features, 100000, True)
    #describe(X_test, lengths_test)

    clf = StructuredPerceptron(verbose=True, max_iter=5)
    print("Training %s" % clf)
    clf.fit(X_train, y_train, lengths_train)
    joblib.dump(clf, 'trunk.pkl')

    import time
    xclf = joblib.load('trunk.pkl')
    start = time.time()
    y_pred = xclf.predict(X_test, lengths_test)
    end = time.time()
    print((end-start)*1000)
    print("Accuracy: %.3f" % (100 * accuracy_score(y_test, y_pred)))

CoNLL format

load_conll

ref : http://larsmans.github.io/seqlearn/reference.html

StructuredPerceptron

joblib